Bioinformatics Seminar
Time: 11AM
Venue: Davis Auditorium and Online
4 November 2025
This is a WEHI only event.treeBayes: An Integrated Framework for Sequential Bayesian Ensemble Learning with Uncertainty Quantification and Explainability
Jinjin ChenWEHI Bioinformatics
treeBayes is an R package enabling reliable machine learning on small-to-large biomedical datasets. It leverages Bayesian ensembling of Random Forest and XGBoost models with principled synthetic data augmentation, validated through multiple distance metrics. The framework quantifies predictive uncertainty by decomposing aleatoric and epistemic components while reporting model variance, prediction disagreement, and confidence intervals. SHAP-based interpretability provides both global feature importance and sample-level explanations for clinical utility. A sequential prior-strength framework enables efficient batch autoregressive updates. Together, treeBayes delivers robust classification with uncertainty quantification, biomarker discovery, and clinically interpretable predictions, addressing critical needs in precision medicine and translational research.